Tingguang Li, Delong Zhu, Max Q.-H. Meng
Despite a large number of specialty mobile manipulators have been deployed in specially designed warehouses, it is still a challenge to accurately estimate the position and orientation of randomly deposited goods and articles. In this paper, we present a hybrid 3DoF pose estimation method, in which camera and Lidar data is fused to accurately estimate the pose of target cubic parcels. We firstly developed a vision-based and a Lidar-based pose estimation method, respectively. Then a hybrid method based on the vision and Lidar methods is proposed to enhance the accuracy and robustness of the pose estimation process. The hybrid method fuses the two outputs from the vision and Lidar methods, employing Kalman filter. In real-world experiments, our real-time system achieves some 1cm mean position error and <1° mean angle error, and shows its significant robustness when applied in our mobile manipulation platform.
@inproceedings{Tingguang:2017, author = {Tingguang Li, Delong Zhu, and Max Q.-H. Meng}, title = {A Hybrid 3DoF Pose Estimation Method based Camera and Lidar Data}, booktitle = {Robotics and Biomimetics (ROBIO), 2017 IEEE International Conference on}, year = {2017}, organization ={IEEE}}